Module simd_tier1 

End-to-end Tier-1 scalar-flow SIMD promotion.

This module deliberately implements only the sealed perfect-ordinal slice: one selected scalar output, one u64 ordinal input that varies inside an owned source lease, pure exact total lane-independent member nodes, and scope-stable scalar broadcasts. The canonical scalar DAG is retained as the oracle and fragment/recovery path; the synthesized register DAG is a parallel execution plan, not a mutation of scalar graph semantics.

Structs

Tier1LeaseStartError

Recoverable lease-start failure. Ownership of the reserved ordinal range is returned to the caller so it can be processed through another scalar path.

Tier1SimdDescriptor

What a Tier-1 SIMD plan promotes: the output, the driving input, the lane shape, the member nodes, and the broadcast inputs.

Tier1SimdExecutor

Per-fiber Tier-1 executor. The scalar and native register kernels share no mutable state with another fiber.

Tier1SimdStats

Counters of a plan’s execution.

Enums

Tier1SimdError

Why a Tier-1 SIMD plan could not be built, started, or advanced; Display renders each with its names.